Deep learning for high performance computing
Are you working with large numerical simulations that take days or months to run, even on a supercomputer? And could you do much more – or even new – science if those simulations ran faster? Then you may want to learn if deep learning could help you accelerate your numerical simulation code.
Many of the traditional workloads on supercomputers are large, numerical simulations. With the increasing success of deep learning in various applications, the SURF Open Innovation Lab together with our university partners has taken the first exploratory steps to discover how deep learning can be used to augment these traditional simulations. Can it be used to accelerate your code? Or can it improve the accuracy of your results, in the same computational time?
In this workshop, we share our experiences, and challenge you to think about opportunities for applying deep learning to accelerate numerical simulations in your field.
PIs & (senior) researchers that use or develop large, numerical simulations.
SURF will give a presentation on a structured approach that allows you to accelerate your numerical simulation with deep learning. In addition, two collaboration partners give a presentation about their specific project and illustrate the SURF approach.
- 14:00 - Improving Deep Learning of large-scale numerical simulations: a structured approach (Caspar van Leeuwen)
- 14:30 - Subgrid turbulence modeling using neural networks (Robin Stoffer)
- 15:00 - Event Generation with the B-UAE (Sydney Otten)